Seeking Alpha
View as an RSS Feed

Ashraf Eassa  

View Ashraf Eassa's Comments BY TICKER:

Latest  |  Highest rated
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    >>You are grasping at straws and your reasoning is flimsy. Linking >>availability of processors with those of chipsets and motherboards is >>just nonsense.

    Well then let's just agree to disagree at this point. I've provided multiple data points from multiple angles and it's all nonsense to you. That's fine.

    I have learned my lesson about trying to share what I know here :-)

    Nov 27, 2015. 10:01 AM | Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    >>As an investor (this is an investment forum) why do I care if smaller >>customers are last in line to get their products? As long as Intel is >>meeting its OEM commitments, and maximize revenue and profits I >>am happy.

    Hey, from an investment standpoint, if Intel is able to meet the 6700K/Skylake demand with older generation Haswell chips at healthy margins (22nm is in great shape), then I don't really care either. I think the recent price cuts in the 4790K, 5820K, and 5930K are really good ways to try to capture the 6700K demand that cannot be satisfied. I hope it works!

    >>"Some of the recent reports about shortages are due to people >>ordering products more than what we predicted, which is obviously a >>good thing, but shortage nevertheless"

    Look, the 6700K/6600K, which are the high end Skylake chips that are in short supply launched in early August 2015. The fact that these chips are in short supply nearly 4 months later is a big red flag to me that says it's more than a "demand greater than expected" problem; it smells like a "oh, dang, we can't build as many of these high performance/high spec chips as we'd like."

    And there's a really, really easy way to verify this. Every Skylake desktop chip needs to be plugged into a motherboard to work. These motherboards contain a "Platform Controller Hub" chip. In the case of Skylake the chipsets are known as the 100-series chipsets.

    These chipsets are built on Intel's high-yielding, very healthy 22nm process.

    If you go on or, there is absolutely no shortage of motherboards based on these 22nm chipsets. There is, however, a shortage of Skylake 6600K/6700K chips.

    If demand were truly "greater than expected" then wouldn't we see "shortages" in both the motherboards/chipsets as well as the CPUs, and not just the CPUs?

    Nov 26, 2015. 10:33 PM | Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    "ephud also agrees with me that the fact you have no chart at all, even without a scale on the y-axis, means you have no basis for making claims of better yields on the A9."

    Which claim are you referring to re: A9? TSMC v.s. Samsung or the hypothetical Intel-built A9 v.s. TSMC/Samsung-build A9 yields?

    Nov 26, 2015. 01:20 PM | Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    "Apple quietly released a 27" iMac with Intel Skylake quadcore processor. It is available right now for shipping."

    You do realize Apple is a very important customer of Intel's (all of the major OEMs are), so Intel will bend over backwards to make sure Apple has enough supply to meet its demand.

    The big/high volume OEMs seem to be getting pretty much what they need, but the smaller guys suffer because the big OEMs are taking up all of the chips.

    "All those pundits here who have been incessantly spreading rumors like Skylake shortage or 14 yield woes, conveniently ignore this."

    You can just insult me directly, there is no need to do it indirectly ;-)

    Anyway, Intel's Bill Holt said himself that 14nm yields aren't where they want them to be and that this has led to a "shortage" of Skylake chips. Listen to the presentation, he literally says this.

    "Nor do they acknowledge strong demand uptick for Intel processors."

    OK, so ask yourself something, 101StockEnthusiast101. If this "shortage" is due to this big, unexpected demand uptick for Intel's processors, then why are the following true:

    1. Intel saw desktop chip volumes drop 15% and notebook chip volumes drop 14% year-over-year last quarter.

    2. Intel is only guiding to a slight sequential increase in revenue from Q3->Q4, which suggests that PC chip sales will still be down substantially year-over-year.

    If this "shortage" were due to an expected uptick in demand, then shouldn't Intel have raised its Q4 guidance at the investor meeting from the levels it first estimated in Q3? Or at the very least, if they couldn't raise guidance because they weren't able to meet demand at least give some indication that channel inventories would be down significantly in Q4 to allow investors/analysts to model in some channel "refill" in Q1?

    Anyway, Intel seems to be guiding to a PC market that's slightly down next year so they don't exactly share your enthusiasm that Skylake demand is far more robust than expected.

    Nov 26, 2015. 01:14 PM | Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    I don't know why I keep replying to you since it is clear that your agenda is simply to prove to the folks who are reading that I have some nefarious anti-Intel agenda and/or am clueless, but here we are.

    "You've stated that Holt's comparison of a Haswell Core M and a Broadwell Core M is meaningless because they're not the same die. You've also implied that Holt is up to something nefarious because his chart doesn't provide a scale on the y-axis."

    I revised my "meaningless" statement to say "it's not saying what you think it's saying." If Intel's yield rate on an 82mm^2 die on 14nm is lower than its yield rate of a 130mm^2 die on 22nm, then this points to a 14nm process that's quite a bit less healthy compared to the 22nm process.

    The lack of a scale on the y-axis does limit the value of the chart, something that ephud agrees with as well.

    "Firstly, why would you be comparing yield on the A9 produced at TSMC, with a "top bin Skylake" which you don't even define? "

    The point I was trying to make is that comparing yield rates of a top bin Skylake (in my mind a 4+2 Skylake at 4GHz base/4.2GHz turbo) and the A9 would be unfair because the Skylake chip is a far more aggressive design.

    "Aren't those not only two different die, of different sizes, but they're also of two different design targets, and from two entirely difference processes from two different companies?"

    Yeah, that was my point. Don't know why you are foaming at the mouth here when we are in agreement.

    Nov 26, 2015. 01:00 PM | Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    Fair enough. What's interesting though if you look at transistor fin images of the foundry FinFET processes, you'll notice that those fins look much more like Intel's 14nm fins than they do Intel's 22nm fins.

    So although we could argue that 16FF+ is TSMC's "first generation" finFET, I don't think it's fair to say that what TSMC/Samsung are producing today (from an xtor performance perspective) are more comparable to Intel's 22nm finFETs than to its 14nm finFETs.

    To see what I am talking about, look at this image of Samsung's 14nm LPE finFET:

    Here are Intel's 22nm and 14nm fins side-by-side:

    Nov 26, 2015. 12:48 PM | Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    If I may provide some color...

    TSMC's 16FF+ is in fact a 2nd generation FinFET. TSMC never went into production on its first generation (just known as 16FF) because it didn't quite meet the performance specifications that its customer(s) wanted.

    The key pitches (transistor gate pitch, minimum metal, etc.) did not change, but TSMC did do quite a bit of engineering work to improve the performance of the actual transistors. What they did exactly hasn't been publicly disclosed, but in its IEDM 2014 paper TSMC said that they were able to achieve a meaningful reduction in capacitance relative to 16FF.

    The issues that Intel is having with 14nm, from what I am told by a source familiar with the matter, lie squarely on the shoulders of the very aggressive minimum metal pitch that the company went for (52nm).

    With respect to the foundries, I am told that Samsung's yields on the A9 chip are significantly worse than TSMC's; the delta was, last I heard, on the order of 2x in favor of TSMC. I suspect TSMC benefited from having a full year of production of 20nm while Samsung skipped over 20nm LPP (which featured a 64nm Mx pitch like the TSMC 20nm technology) in favor of 14nm LPE/LPP.

    Hope this helps.

    Nov 26, 2015. 11:32 AM | 3 Likes Like |Link to Comment
  • Report: Apple to begin using OLED iPhone displays in 2018 [View news story]

    Nope, no negative implication for corning. Still need a cover glass for the phone :)

    Nov 26, 2015. 06:59 AM | 3 Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    I think TSMC's A9 yields may be higher than top bin Skylake yield, but I think if Intel were building the A9, there's a good chance that Intel 14nm yields would be better.

    Nov 25, 2015. 06:37 PM | Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    I would bet a lot of money that if Intel is building an SoC for LG, it will not integrate Intel modem IP, at least not for something coming in 2016.

    Nov 25, 2015. 01:56 PM | Likes Like |Link to Comment
  • The New Apple iPad Pro Appears Targeted At Amateurs [View article]
    Henry 3 Dogg

    Sorry, but this is just plain wrong. RISC/CISC has nothing to do with it. The ability to emulate another ISA satisfactorily really comes down to the quality of the emulation layer + the speed of the chip doing the emulating.

    I think the A9X would have a very, very hard time emulating x86 at a speed that Mac users would find acceptable.

    Don't count on Apple ditching Intel/x86 in the Mac anytime soon, IMO. Intel's PC chips are first rate and the company has a wide variety of SKUs that can serve all of Apple's Mac needs.

    Nov 25, 2015. 12:16 PM | 1 Like Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    "Again, I have serious doubts. If TSMC gets to 10 nm first and Apple releases the A10/A10X in a timely fashion, even Intel's top of the line Core i7 CPU will be disadvantaged performance wise and that would be disastrous as the A series has outstanding power consumption characteristics along with availability at a superior price."

    A10 will be manufactured on TSMC 16FF+, not 10nm (and not at Intel/Samsung, either).

    Nov 25, 2015. 11:05 AM | Likes Like |Link to Comment
  • Intel: Reasons To Curb Your Enthusiasm, As 2016 Looks Like Another Yawner [View article]
    "So they are now upping Capex to 10 billion, but they don't say for what."

    They did, but you need to have actually paid attention to the presentations to have figured this one out.

    Nov 24, 2015. 07:53 PM | 7 Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    >>It's also telling that you keep pointing to old data from August 2014, >>rather than looking at current data from Nov 2015.

    They don't mark PRQ dates on the latest chart, sorry. Gave you the best answer that I could given my understanding of the question that you posed.


    p.s. the actual realized cost structures of Intel's processors on 14nm have already answered the question.
    Nov 24, 2015. 04:06 PM | Likes Like |Link to Comment
  • LG Nuclun 2 Confirms Intel's Foundry Superiority [View article]

    Slide 37.

    Nov 24, 2015. 03:48 PM | 1 Like Like |Link to Comment